Who she is
Affiliation: Charles Edward Wilson Professor Emerita, Harvard Business School (unchanged — employer check run 2026-09-21; El País identifies her as professor emeritus, 2025-12-15); author In the Age of the Smart Machine (1988), The Support Economy (with James Maxmin, 2002), The Age of Surveillance Capitalism (2019). Working on a 2026 book update focused on AI's role in surveillance capitalism (El País, 2025-12-15).
One-line position: AI is the operational core of a new economic logic — "surveillance capitalism" — that extracts behavioral data as raw material, manufactures prediction products, and sells certainty about future human behavior; the threat is to democracy and individual autonomy, not just privacy. New since v1: "AI is surveillance capitalism continuing to evolve and expand," with a sharpened line on tech oligarchs' influence over political leadership.
Discipline & technical bet
Social theorist and political economist, not a technologist — her "technical bet" is conceptual: that the business model, not the model architecture, is the correct unit of analysis for AI. The live wager: generative AI gets absorbed into the extraction-prediction loop by default (ad-funded assistants, behavioral surplus from conversation data) rather than escaping it via subscriptions and enterprise rents. The 2026 book is that wager committed to print. She makes no capability claims and takes no position on timelines — deliberately orthogonal to the scaling debate.
Key claims (Says)
- Empirical / structural: a new accumulation logic — surveillance capitalism — emerged at Google ~2001–2004, was perfected as advertising targeting, and has spread across sectors (Meta, Amazon, Microsoft, insurers, automakers, ed-tech). Held — restated 2025-12: extraction has "expanded nearly unrestricted" since 2019 (El País, 2025-12-15).
- Empirical: behavioral surplus — data exhaust beyond service improvement — is the raw material; unilaterally claimed, not exchanged. Held — analytical frame, still the load-bearing concept; "very few things left in this world that we can do without contributing to it" (El País, 2025-12-15).
- Empirical: surplus is processed through machine intelligence into prediction products sold in behavioral futures markets (advertisers, insurers, campaigns, employers). Held — the conversational-AI ad market now forming is a direct new instance (OpenAI ads test, 2026-01-16).
- Empirical: monitoring → prediction → modification — "instrumentarian power" that nudges and herds rather than coerces. Held (directionally) — feed-shaping and nudge architectures intensified; at-scale causal measurement remains contested (weak spot retained).
- Normative: an illegitimate accumulation form — morally analogous to enclosure of common land — because it removes human experience from individual sovereignty without negotiation. Held — unchanged.
- Empirical / political: surveillance capitalism is incompatible with democracy at scale — asymmetric knowledge and power. Held and escalated — her 2025-12 framing adds the oligarchic turn: named concern about tech billionaires' direct influence on political leadership (El País, 2025-12-15), which the 2025–26 US platform-politics era furnished with daily examples.
- Normative: the remedy is prohibition of extraction and prediction-product markets, not "data ownership" (which embeds the trade in market logic) — organ-market analogy. Held — unchanged, and still operationally under-specified (weak spot retained).
- Empirical: AI as deployed is the operational core of surveillance capitalism — not a neutral capability later misused. Drifted — half the story now. The extraction loop is being fitted to chatbots on schedule (ads in ChatGPT, tested from 2026-01-16), but the economic center of frontier AI in 2024–26 moved to capability rents — subscriptions, enterprise API, and a trillion-dollar infrastructure buildout (see Andreessen Machine Age entry) — none of which is behavioral-futures revenue. AI's political economy has TWO engines now; her frame prices only one of them.
Notable predictions — with falsifiable checks
- (2019) The logic spreads from advertising into insurance, employment, education, healthcare, public administration. Held — telematics, hiring algorithms, ed-tech monitoring, health-data partnerships all expanded through 2026; restated by her 2025-12. Check: any major sector where extraction-based pricing/scoring is rolled back by regulation before 2028.
- (2019) Increasing pursuit of behavioral modification, not just prediction. Held (directionally) — recommendation and nudge architectures intensified. Check remains weak (measurement problem) — flag: define a concrete observable next cycle or retire the grade.
- (2019) Democratic institutions lack epistemic parity and will struggle to respond. Held — EU DSA/DMA/AI Act address surfaces, not the prediction-product model; US federal privacy law still absent; the 2025–26 preemption push moved the venue fight away from her preferred remedies. Check: does any DSA Art. 40-style researcher-access regime produce systemic-risk findings with enforcement teeth by 2028?
- (2021, NYT) Generative AI will be absorbed into the surveillance-capitalist logic by default unless explicitly prevented. Held — OpenAI began testing ads in ChatGPT in the US (CNBC, 2026-01-16; OpenAI announcement), including sponsored-agent formats — the ad-funded assistant she predicted, arriving five years later. Check: share of frontier-lab revenue from advertising/attention products at end-2027 — if it stays marginal (<10%), the absorption claim weakens against the capability-rent counter-story.
- (2019–2024) "Epistemic inequality" will deepen without prohibition. Unfalsifiable as stated — rewritten with observable (rule: no observable, no grade): proxy check = researcher/regulator access to platform and model internals mandated and exercised (DSA Art. 40 data access, AI-incident reporting mandates) by 2028; if access regimes exist and function, the asymmetry claim needs restating.
Revealed behavior (Does)
- Writing, not organizing: the observable output is a 2026 book update on AI (El País, 2025-12-15) — she doubles down on the long-form conceptual register rather than policy shops, litigation, or movement-building.
- Keeps lecturing into elite venues (Harvard STS, EPFL, European Parliament circuit, Lavin speaking bureau) — the theory travels through institutions, not through a mass audience product.
- Uses the tools she indicts (the El País interview ran on Google Meet, an irony she acknowledges) — revealing her actual claim: exit is individually impossible, which is the point.
- Has not taken board seats, advisory roles, or fund positions in tech — the no-commercial-exposure profile is a durable behavioral fact, re-verified this cycle, not just a credential.
Feels
Urgency shading into vindication-weariness — the 2019 warnings became the 2026 furniture and the prohibition she asked for never came. What she wants to be true: that naming a logic precisely is the first step to outlawing it, as it once was for child labor and organ markets.
Hears
Critical political-economy lineage (Polanyi, Arendt, Durkheim), European regulators and parliamentarians, critical-tech scholars and journalists; systematically thin on working technologists and on the capability-safety research world — by design, not oversight.
Sees
The longest baseline on the bench: she was writing about computer-mediated power in 1988, so she sees the AI wave as chapter three of one forty-year process — a vantage that catches business-model continuity everyone treating 2022 as year zero misses.
Incentive map
Re-derived this cycle (rule: no status is exempt). Emerita position, book royalties, lecture fees; the 2026 book gives her a near-term commercial stake in the thesis staying alarming — standard author's bias, disclosed by the publication calendar itself. Reputational lock-in is the larger force: "surveillance capitalism" is her coinage and her legacy; a world where AI's center of gravity is capability rents rather than prediction markets is a world she is structurally slow to describe. Cannot easily say: "the extraction model turned out to be the lesser problem." Rule-25 discount: low — but apply the frame-inertia discount instead of a financial one.
Theories aligned with
- Surveillance capitalism (her term) — primary frame
- Critical AI ethics — adjacent, shares concern with deployed harms and political economy (no file — theory not written up)
- Critical political-economy lineage — Polanyi (The Great Transformation), Arendt (totalitarianism / power), Durkheim (division of labor) — explicitly cited
- Distinct from but partially overlapping with Varoufakis's techno-feudalism (different diagnosis, similar urgency)
- Adjacent to post-labor economics on the asymmetry question, not on the redistribution mechanism
What she's reacting against
- Framing Big Tech as ordinary capitalist firms or as a "privacy" problem — argues neither captures the new mode of accumulation
- Tech-optimist framing of personalization, targeting, and recommendation as user-benefit services — argues these are extraction-then-prediction loops
- "Data as the new oil" metaphor — argues it's worse: oil is found, behavioral surplus is expropriated from human experience
- Regulatory framings (GDPR, US sectoral privacy law) that treat the problem as consent / disclosure rather than as an illegitimate accumulation regime
- Conflation of her critique with Varoufakis-style techno-feudalism — argues surveillance capitalism is still capitalism, just a mutant form
- NEW: the oligarchic merger of platform wealth and political power — the 2025–26 addition to her target list (El País, 2025-12-15)
Where she overlaps / splits (with Rich)
- Overlaps with Varoufakis on the diagnosis that Big Tech is structurally different from prior capitalism; splits on whether it's still capitalism (Zuboff: mutant capitalism) or post-capitalism (Varoufakis: techno-feudalism) — the ad-funded-chatbot turn is evidence both can claim
- Overlaps with Gebru on prioritizing present, deployed harms and Big Tech political economy; splits on emphasis — Gebru centers representational / labor harm and TESCREAL critique; Zuboff centers the prediction-market extraction logic
- Overlaps with Acemoglu on institutions-matter framing; splits on whether reform inside the current regime suffices — Zuboff wants the extraction logic prohibited, not regulated
- Splits sharply with Andreessen, Thiel, Altman on whether the current regime is "capitalism functioning normally" — and Altman's ads test makes OpenAI a direct new exhibit in her case
- Splits with Diamandis, Blundin, Ismail on whether the abundance frame addresses the asymmetric-power problem (it doesn't, in her view)
- Splits with Yudkowsky, Bengio, Amodei on framing: x-risk is downstream of, or orthogonal to, the already-operative threat of behavioral expropriation and democratic erosion
- Overlaps with Toner on the need for external accountability; splits on whether the problem is lab self-regulation (Toner) or the underlying business model (Zuboff)
- Less engaged with Brynjolfsson, Cowen on labor-economics specifics — she argues the wage / employment frame misses the deeper expropriation
Track record
- In the Age of the Smart Machine (1988) — early, prescient analysis of computer-mediated work and "informating"; durable citation in workplace-IT and STS literature
- The Support Economy (2002, with James Maxmin) — "distributed capitalism" frame; less commercially influential, but anticipated platform critique
- The Age of Surveillance Capitalism (2019) — moved her coinage into common policy vocabulary; widely taught, 20+ translations
- "Surveillance Capitalism or Democracy? The Death Match of Institutional Orders" (Organization Theory, 2022) — the academic consolidation of the democracy argument, bridge to the 2026 book
- Influence on EU regulatory discourse (DSA, DMA, AI Act) real but indirect; legal codification partial — her own prediction about institutional struggle, confirmed on herself
- Pattern: long-form conceptual framings that age well analytically; forecast register thin, but her one sharp genAI call (2021: absorption by default) landed in January 2026
Empirical vs normative
- Empirical: distinct accumulation logic with documented mechanisms; cross-sector spread; genAI absorption now evidenced by the ChatGPT ads test; the modification claim remains the weakest empirical link
- Normative: the logic is illegitimate; democracy requires that human experience not be raw material; remedy is prohibition, not consent or ownership
- Conflict note (CONVENTIONS rule 25): minimal financial conflict, re-verified 2026-09-21; the operative discount is frame-inertia plus a forthcoming book's thesis-stake, not portfolio exposure
Weak spots / open questions
- "Prohibition of behavioral futures markets" remains operationally under-specified — what exactly is prohibited, by what mechanism, with what evidentiary standard? Seven years on, no draft statute carries her remedy
- The "mutant capitalism vs post-capitalism" line is sometimes drawn ambiguously; critics argue she wants the rhetorical force of "post-capitalism" without the analytical commitment (Morozov, 2019)
- Largely silent on the capability side of AI (alignment, reasoning, agency) — orthogonal to Bengio / Amodei; a feature for clarity, a bug for comprehensive AI policy, and increasingly costly as capability rents become AI's second economic engine
- Monitoring→prediction→modification is theoretically clean but empirically contested at scale — modification vs selection effects still unresolved; next review should attach an observable or retire the grade
- Underweights the Mostaque / open-source argument that decentralized AI infrastructure could break the extraction logic
- Less attention to non-Western surveillance-state architectures (China's social credit, state surveillance) — primarily a critique of US Big Tech political economy
- The 2026 book is the test: if it engages the capability-rent economy and the oligarchic turn analytically, the frame is alive; if it restates 2019 with AI examples, the persona ages to legacy status
Rich's take
- (your synthesis here)
Delta log
2026-09-21 — v1→v2 migration + validation (weekly batch)
- Grades: 10 Held / 1 Drifted / 0 Wrong (one prediction rewritten as unfalsifiable-with-proxy per bench rule).
- Held: accumulation-logic spread ("expanded nearly unrestricted," El País, 2025-12-15); behavioral-surplus frame; prediction-products/futures-markets (new instance: conversational-AI ads); modification (directional); illegitimacy (normative, unchanged); democracy-incompatibility (escalated to oligarch framing); prohibition remedy (unchanged); sector-spread prediction; modification prediction; institutional-struggle prediction; genAI-absorption prediction — the standout: OpenAI began testing ads in ChatGPT (CNBC, 2026-01-16), landing her 2021 call almost exactly.
- Drifted: AI-as-operational-core — the extraction loop is real and growing (ads test), but 2024–26 frontier AI's economic center is capability rents and infrastructure capex, a second engine her frame doesn't price; the claim is now half of a two-engine story.
- Most surprising delta: her least-respected prediction aged best — the 2021 "genAI absorbed by default" call, dismissed as reflexive pessimism when subscriptions looked like the model, was confirmed by OpenAI's own ads announcement in January 2026.
- Employer check (mandatory): no change — Professor Emerita, HBS; no boards, funds, or advisory roles found. New fact: 2026 AI-focused book update in progress (El País, 2025-12-15) — wake trigger set.
- Tier: annual — concepts move on book-length timescales and her role is stable; wake triggers cover the fast paths (book lands, prohibition enacted anywhere, ad-funded AI dominance, repositioning).
- Confidence unchanged at medium.
Sources
- 2025–26: El País interview — "AI is surveillance capitalism continuing to evolve and expand" (2025-12-15, P. Nagovitch; reprint) · CNBC — OpenAI to begin testing ads on ChatGPT in the US (2026-01-16) · OpenAI — "Testing ads in ChatGPT"
- Books: In the Age of the Smart Machine (Basic Books, 1988); The Support Economy (with James Maxmin, Viking, 2002); The Age of Surveillance Capitalism (PublicAffairs, 2019); AI-focused update announced for 2026
- Peer-reviewed: "Big Other: Surveillance Capitalism and the Prospects of an Information Civilization" (Journal of Information Technology, 2015) · "Surveillance Capitalism or Democracy? The Death Match of Institutional Orders" (Organization Theory, 2022)
- Op-eds: "You Are Now Remotely Controlled" (NYT, 2020-01-24); "The Coup We Are Not Talking About" (NYT, 2021-01-29)
- Talks / interviews: The Ezra Klein Show (NYT, ~2021); Harvard STS and EPFL campus lectures; European Parliament addresses; The Social Dilemma (Netflix, 2020)
- Critique: Morozov, "Critique of Techno-Feudal Reason" (NLR, 2019)
migrated v1→v2 2026-09-21 · backup: _archives/shoshana-zuboff.html.bak-20260921 · converts-from: people/shoshana-zuboff.md · AI & Society domain